This breakdown of New Doom's rendering is extremely great http://www.adriancourreges.com/blog/2016/09/09/doom-2016-graphics-study/ …
-
-
Replying to @AmazingThew
Approach makes sense but I have so many questions about optimization. 1300 draw calls, 50 render targets, *80* texture taps for blur??
3 replies 1 retweet 0 likes -
Replying to @AmazingThew
True circular bokeh is difficult without a lot of taps.
1 reply 0 retweets 0 likes -
Replying to @cmuratori
Sure; my point is more if I put an 80-tap blur in my game that'd be like 40% of frame budget for one effect,
1 reply 0 retweets 0 likes -
Replying to @AmazingThew
It doesn't necessarily work like that. Taps are not uniformly expensive. Nearby taps hit the cache, etc.
1 reply 0 retweets 0 likes -
Replying to @cmuratori @AmazingThew
So you may be surprised how fast 80-tap blurs go on modern hardware :)
1 reply 0 retweets 0 likes -
Replying to @cmuratori
I didn't think cache covered anywhere near a radius that large? I guess I've never seen actual numbers for it...
1 reply 0 retweets 0 likes -
Replying to @AmazingThew
24k is a lot of cache if all you're doing is reading from the same (half-res!) texture.
1 reply 0 retweets 0 likes -
Assuming a 64 byte cache line, you could read 384 completely disparate pixels from the cache (bullshit analysis, but still)
-
-
Replying to @cmuratori @AmazingThew
Anyway the point is that 80 taps into the same texture in the same place is very different than, say, sampling 80 textures.
1 reply 0 retweets 0 likes -
Replying to @cmuratori
Certainly. I guess I'm mostly just underestimating how much faster a 980 runs compared to my 770
0 replies 0 retweets 0 likes
End of conversation
New conversation -
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.